We studied single-neuron activity in the prefrontal cortex (PF) while a monkey performed a task according to two different rules, termed conditional and spatial. The monkey viewed a video screen, and its task required a hand movement in response to the dimming of a light spot. There were four light spots on the screen: right, left, up, and down from the center. Only one of the four spots dimmed, and the degree of dimming was slight. Accordingly, the monkey needed to foveate the "correct" light spot to detect the dimming. A visual cue indicated which of the four light spots would be deemed correct and, thus, would dim on each trial. The sequence of events was as follows: a fixation spot appeared at the center of the screen; then, a cue appeared twice at one of the four potential target locations; then, the four target spots appeared; and, finally, one of them dimmed. Except for the color of an initial fixation point, the cues, their locations, and other events were identical for the conditional and spatial rules. The rules differed in one essential way. For the conditional rule, nonspatial attributes of the visual cue indicated which of the four light spots would dim, and the cue's location was irrelevant. For the spatial rule, the cue's location determined the correct target on that trial. The light spot at the location of the cue always dimmed, regardless of which cue appeared there. Our sample included 221 PF neurons showing significant task-related activity modulation, distributed among dorsal, dorsolateral, and ventral PF regions. Between one-third and one-half of the sample in each of those regions showed statistically significant activity differences that could be attributed to the rule. Selectivity for cues and/or their locations was common. However, there was no significant regional segregation of such selectivity. These data support the hypothesis that PF plays a role in the guidance of behavior according to previously learned rules.